What's this about?
Mollick names the gap: "multiplayer AI" — many people in an organization using AI *together* — is still stuck at AI-as-a-person-in-your-group-chat, while single-player AI races ahead.
What exactly is missing?
Shared context, shared memory, permissions across people, an AI that knows the whole team's state — the collaboration layer. Every tool today assumes one human per session.
Why is this worth noting?
Organizations run on collaboration, so single-player gains hit a ceiling — and whoever cracks the team-shaped product owns a mostly empty market. Notable that the sharpest adoption researcher calls this, not model quality, the open frontier.